OpenWork – An Open‑Source Claude Cowork Alternative

OpenWork is a modern desktop application written in TypeScript and Rust (via Tauri) that wraps the OpenCode engine into a clean, guided, product-like workflow. The project is a direct answer to developers’ frustration with OpenCode’s CLI‑centric tooling and offers an intuitive UI for typical knowledge‑worker scenarios.

Why OpenWork?

• Extensible – Install and manage OpenCode plugins, custom skills, and template workflows via a native skill manager.
• Auditable – Every run displays an execution plan, SSE updates, and permission requests, so you can see exactly what the AI is doing.
• Permission‑Managed – Prompted once‑or‑always access to privileged actions keeps your data safe.
• Local & Remote – Run OpenCode locally (opencode serve) or point to a remote server. All modes expose the same UI.
• Product‑like – A fully‑featured app with folder pickers, progress bars, and reusable templates removes the need to learn command‑line toolchains.

Project Overview

Key files and concepts in the repository:

Quick Start Guide

Prerequisites

1. Node & pnpm – npm i -g pnpm
2. Rust toolchain – curl https://sh.rustup.rs -sSf | sh
3. OpenCode CLI installed (opencode on PATH)

Install and Run

# Clone the repo
git clone https://github.com/different-ai/openwork
cd openwork

# Install dependencies
pnpm install

# Start locally (desktop)
pnpm dev

# Or start only the web UI (useful for remote sessions)
pnpm dev:web

You can download pre‑built .dmg releases from the Releases page or build your own with pnpm build:web.

Architecture – How the Pieces Fit Together

1. Host Mode – OpenWork spawns a local OpenCode server (opencode serve --hostname 127.0.0.1). The UI connects via the @opencode‑ai/sdk/v2/client.
2. Client Mode – Pointing to an existing OpenCode server URL exposes the same session, prompt, and SSE APIs.
3. UI Layer – React components handle workspace selection, prompt submission, and render the execution timeline.
4. Skill Management – The Skills tab lists all .opencode/skill/ folders, allows installation from the OpenPackage registry, or manual import.
5. Permission Flow – OpenWork intercepts OpenCode permission requests and presents modal dialogs to the user.

Core Features Walk‑through

• Workspace Picker: Open a folder that becomes the working directory for OpenCode.
• Execution Plan: Visual timeline of the AI’s to‑do list.
• Live Streaming: Real‑time updates via Server‑Sent Events.
• Templates: Save common workflows for quick re‑execution.
• Skill Manager: Install or remove OpenCode plugins with a button click.
• Audit Trail: Each run logs the prompt, plan, and outcome in a JSON log.

Extending OpenWork

OpenWork embraces the plug‑in model of OpenCode. Add a new skill by creating a folder inside opencode/skill/ or using opkg install <package>. The opencode.json file in either the project root or global config folder determines which plugins are active.

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-wakatime"]
}

Edit this file manually or use the UI to toggle plugins. The change takes effect on the next server start.

Security & Privacy

• Opens locally only (127.0.0.1) by default, protecting against remote tampering.
• Masks raw model logs and sensitive tool metadata unless explicitly requested.
• Permission prompts allow granular control over what the AI can access in your environment.

License & Community

OpenWork is MIT‑licensed. Contributions are welcome—feel free to open issues or pull requests on GitHub. The project has an active community of developers and knowledge workers building productive AI tooling.

Bottom Line

OpenWork turns the powerful, yet developer‑centric OpenCode engine into an approachable, audit‑friendly desktop workflow. For knowledge workers who prefer a GUI over a terminal, this open‑source project fills a vital niche—providing a clean interface, template reusability, and permission safety, all while keeping the underlying AI logic fully transparent and modifiable.

Ready to give agentic work a product‑like touch? Clone the repo, install the dependencies, and start your first session today!